It is thought to be an important brain function for us to modulate our cognitive state depending on recognition of interactive agents. From our behavioral experiment, we have shown that when subjects believed their interactive agent as a human being in a competitive game, specific behavioral tendency could be observed in comparison with a case that they believed their opponent as a computer and this tendency of subjects with autism spectrum disorder (ASD; they are thought to have some problems in social interaction) was different from that of subjects without ASD. We consider that this tendency is a reflection of cognitive modulation depending on recognition of interactive agents. To explain computational theory of this tendency, we propose a computational model that consists of change detection and state space switching evoked by the change of environmental nature. From this model, we reproduce the result of behavioral experiment by a computer simulation and try to discuss our computational model is useful to understand about the brain function for frequent human social interaction.